Under the "dual carbon" strategy, the low-carbon supervision of zero-carbon computing center construction faces a paradigmatic dilemma—traditional supervision models remain confined to "carbon emission monitoring" and fail to achieve a closed loop of "carbon emission governance." Drawing on the resource-based view and dynamic capabilities theory, this paper proposes a four-step innovative pathway of "digitalization of rules—three-dimensional perception—intelligent analysis—closed-loop disposal," and elucidates the organizational logic of the AI supervision large model as a core enabling mechanism. Using the AI supervision large model and digital supervision platform of a leading engineering consulting firm as a case study, we demonstrate the managerial feasibility of this pathway. The findings indicate that the AI supervision large model can upgrade ex-post carbon accounting to in-process carbon governance, thereby providing a replicable strategic framework for engineering consulting firms to develop new zero-carbon technology service businesses.